Reference
dataset
(Map)
Initial registration
Define RoI
Segment Select layers
Edge extraction
Rasterise with
attributes
— +
Find
corresponding
polygons
I
Find
corresponding
points
]
FindXYof HW 2D Warp
common points
Find Z of
common points
Orthoimage Validation
=
Camera model
re Produce
Figure 1. A strategy for image to map registration.
The general stages used in extracting polygons from the map
are smoothing, segmentation, edge enhancement, edge
thinning and removal of small polygons. It must be
remembered that for the process of registration only a small
proportion of features are needed and the rejection of some
polygons is not a problem.
2.2 Matching the polygons
Polygon matching is required at two levels. First
corresponding polygons must be established and then the
edges of these must be matched on a point by point basis.
To carry out the first stage an initial, approximate
correspondence must be established. A number of techniques
are available. Alternatives which are available include the
manual selection of 2 to 4 control points or an automatic
identification of position and orientation by matching large
polygons, possibly embedded into an image pyramid. The
first two methods have been used but it is thought that the
automatic method may not be robust in all circumstances
using only a single layer.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
The basic techniques of matching polygons is adapted from
Abbasi-Dezfouli and Freeman, (1994). Polygons are
characterised by a number of parameters such as shape and
area. Shape is defined by dimensions parallel to defined axes
and orientation and also by the chain code method described
by Abbasi-Dezfouli and Freeman. The initial translation
and azimuth must be fixed by first defining a few polygons
which have good matches based on a first pass through the
selected points. An iterative approach then allows
corresponding polygons to be identified. A large number of
polygons are not necessary but it is important that they are
distributed in a suitable pattern over the image. The method
is described in detail in section 3.
Once established he corresponding polygons must then be
exactly matched in order to extract conjugate points. A
method of dynamic programming developed at UCL is one
method of doing this, (Newton et al, 1994). The perimeter
of the feature is followed and a best fit obtained. The
matching of the edges uses the method first developed by
Maitre and Wu (1989). Costs are determined by a number of
measures relating the predicted edge pixel position projected
into the map and the edge pixel under consideration. The
difference in gradient direction between the map boundary
pixel and the edge pixel under consideration are used as costs.
The method also allows the detection of changes between the
two polygons which may represent true change or an error in
detection, in either case such points will not be selected as
conjugate. The technique makes allowance for the fact that
the image may be distorted due to terrain effects or geometric
effects from the camera or sensor.
Other techniques are under investigation such as recognition
of corners.
2.3 Transformation
Once the corresponding points are selected then the required
transformation can be computed. If absolute orientation of a
pair of aerial photographs are under consideration then three
dimensional reference data must be available. Model co-
ordinates can be found using the co-ordinates of one image,
the relative orientation elements and an approximate surface
model to predict the area on the second photograph within
which the conjugate point lies. The exact position can be
found by stereo matching. If finite element modelling was
used to find the relative orientation then the output from this
could be used to determine the conjugate co-ordinates. If
only a single image is to be matched then a two dimensional
transformation will be used or a space resection carried out.
3. REGISTRATION OF FIELD BOUNDARIES ON
AERIAL PHOTOGRAPHY TO 1:10 000 MAP DATA
The first example is close to a fully automated system in that
polygons are selected in an entirely automatic manner from
the image and from the map and the method of polygon
matching described by Abbasi-Dezfouli and Freeman,
(1994) is used. Furthermore the two dimensional conjugate
points derived from the image are given a height value so that
a full three dimensional transformation can be carried out.
The method is described by Morgado (1996).
An aerial photograph of the Isle of Wight was scanned using
the Sharp JX-600 scanner at 600dpi resulting in a -26Mb
image. The scale of the photography is 1:11 600 and the
pixel size of the resulting digital image is 0.50m. The
